Commits

Breadcrumbs

History for

onr0.9

User selector

All users

All time

Commit History

Commits on Jun 2, 2016

• Update copyright for 3p/tf/python.

  Show description for 0cf9ed3
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  0cf9ed3

  Copy full SHA for 0cf9ed3

Commits on May 27, 2016

• Features and bugfixes in TensorFlow RNN.

  Show description for 3c402f5
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  3c402f5

  Copy full SHA for 3c402f5

Commits on May 24, 2016

• Add support for arbitrarily nested tuples for RNN state.

  Show description for ae5c66e
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  ae5c66e

  Copy full SHA for ae5c66e

Commits on May 9, 2016

• Add support for RNNCell to read and return states as tuples,

  Show description for 374e673
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  374e673

  Copy full SHA for 374e673

Commits on Apr 11, 2016

• Fix RNN performance bug. + Additions to rnn benchmarks & benchmarks.py.

  Show description for eb161ec
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  eb161ec

  Copy full SHA for eb161ec

Commits on Apr 4, 2016

• Change rnn benchmarks to use variables and initialize only once

  Show description for 5a0e8fb
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  5a0e8fb

  Copy full SHA for 5a0e8fb

Commits on Apr 3, 2016

• Disable optimizations when running benchmarks.

  Show description for 784cb15
  

  keveman

  authored and

  tensorflower-gardener

  committed
  784cb15

  Copy full SHA for 784cb15

Commits on Mar 25, 2016

• Bugfix in RNN python benchmark

  Show description for 4ab4b3c
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  4ab4b3c

  Copy full SHA for 4ab4b3c

Commits on Mar 22, 2016

• Use tf.group in rnn benchmark.

  Show description for 733321f
  

  yuanbyu

  authored and

  tensorflower-gardener

  committed
  733321f

  Copy full SHA for 733321f

Commits on Mar 19, 2016

• Convert RNN benchmarks to new Benchmark test & runner.

  Show description for 9da6f8c
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  9da6f8c

  Copy full SHA for 9da6f8c

Commits on Mar 15, 2016

• Remove use of conitional pass-through from dynamic_rnn in favor of just

  Show description for b10c65c
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  b10c65c

  Copy full SHA for b10c65c

Commits on Mar 11, 2016

• Fix dependencies bugs

  Show description for 56f1d64
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  56f1d64

  Copy full SHA for 56f1d64

Commits on Mar 9, 2016

• Remove the requirement for sequence_length input to dynamic_rnn.

  Show description for e638881
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  e638881

  Copy full SHA for e638881

• Further improvements for training long sequences. Here is the current performance of rnn.dyanmic_rnn() on various sequence lengths. The last column is the time for each time step. It stayed as a co…

  Show description for 204400d
  

  yuanbyu

  authored and

  tensorflower-gardener

  committed
  204400d

  Copy full SHA for 204400d

Commits on Mar 7, 2016

• Make RNN calls require a known input depth (via shape inference).

  Show description for cc62d99
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  cc62d99

  Copy full SHA for cc62d99

Commits on Mar 4, 2016

• More work on GPU/CPU memory swaping. Trigger the memory swapping only when the "in-use" memory reaches a given threshold. Added more bechmarks.

  Show description for 76af3d9
  

  yuanbyu

  authored and

  tensorflower-gardener

  committed
  76af3d9

  Copy full SHA for 76af3d9

Commits on Mar 3, 2016

• Replacing use of legacy graph_util.pin_to_cpu with device('/cpu:0')

  Show description for a8a1370
  

  David G. Andersen

  authored and

  tensorflower-gardener

  committed
  a8a1370

  Copy full SHA for a8a1370

Commits on Mar 1, 2016

• Added a benchmark for the performance of dyanmic rnn with memory swapping. For the large model, the overhead is roughly 35%. More improvements over this base-line implementation are coming. But if…

  Show description for 384480d
  

  yuanbyu

  authored and

  tensorflower-gardener

  committed
  384480d

  Copy full SHA for 384480d

Commits on Feb 29, 2016

• Bugfix to dynamic_rnn: restore some shape information when possible during loop.

  Show description for 1827f66
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  1827f66

  Copy full SHA for 1827f66

Commits on Feb 22, 2016

• Add a few more special GPU kernels for int32 and bool.

  Show description for f0c5277
  

  yuanbyu

  authored and

  tensorflower-gardener

  committed
  f0c5277

  Copy full SHA for f0c5277

Commits on Feb 11, 2016

• Get rid of some import cruft.

  Show description for 7760ce5
  

  josh11b

  authored and

  tensorflower-gardener

  committed
  7760ce5

  Copy full SHA for 7760ce5

Commits on Feb 10, 2016

• Add fully dynamic RNN op. Unofficial version (not yet documented).

  Show description for 45197e0
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  45197e0

  Copy full SHA for 45197e0

Commits on Feb 7, 2016

• Merge changes from GitHub, clean up linter errors, fix dependencies test.

  Show description for 59f1eba
  martinwicke

  committed
  59f1eba

  Copy full SHA for 59f1eba

Commits on Feb 4, 2016

• Merge nodes do not currently forward shapes correctly. Set statically the

  Show description for cfbd9f9
  

  A. Unique TensorFlower

  authored and

  Vijay Vasudevan

  committed
  cfbd9f9

  Copy full SHA for cfbd9f9

Commits on Feb 3, 2016

• tf.nn.rnn(..., sequence_length) now conditionally selects states and output.

  Show description for 50a8628
  

  ebrevdo

  authored and

  keveman

  committed
  50a8628

  Copy full SHA for 50a8628

Commits on Jan 30, 2016

• Use cc_binary rather than cc_library to reduce size of native library in APK from 5.5mb to 3.2mb (compressed).

  Show description for 8a59748
  

  A. Unique TensorFlower

  authored and

  Vijay Vasudevan

  committed
  8a59748

  Copy full SHA for 8a59748

Commits on Jan 28, 2016

• Breaking change in TF RNN python api: Return the final state instead of the

  Show description for fea55e1
  

  ebrevdo

  authored and

  Vijay Vasudevan

  committed
  fea55e1

  Copy full SHA for fea55e1

Commits on Jan 25, 2016

• Running our linter on a lot of files.

  Show description for 9b70316
  

  Vijay Vasudevan

  committed
  9b70316

  Copy full SHA for 9b70316

Commits on Dec 7, 2015

• TensorFlow: upstrea changes from git.

  Show description for cd53f3c
  

  Vijay Vasudevan

  committed
  cd53f3c

  Copy full SHA for cd53f3c

Commits on Dec 4, 2015

• TensorFlow: upstream changes to git

  Show description for 54a644f
  

  Vijay Vasudevan

  committed
  54a644f

  Copy full SHA for 54a644f

Commits on Dec 1, 2015

• TensorFlow: upstream changes to git

  Show description for 795f35d
  

  Vijay Vasudevan

  committed
  795f35d

  Copy full SHA for 795f35d

Loading